diff --git a/mlir/include/mlir/Dialect/GPU/GPUOps.td b/mlir/include/mlir/Dialect/GPU/GPUOps.td
--- a/mlir/include/mlir/Dialect/GPU/GPUOps.td
+++ b/mlir/include/mlir/Dialect/GPU/GPUOps.td
@@ -86,6 +86,55 @@
   }];
 }
 
+def GPU_SubgroupIdOp : GPU_Op<"subgroup_id", [NoSideEffect]>,
+    Arguments<(ins)>, Results<(outs Index:$result)> {
+  let description = [{
+    Returns the subgroup id, i.e. the index of the current subgroup within the
+    workgroup.
+
+    Example:
+
+    ```mlir
+    %sgId = gpu.subgroup_id : index
+    ```
+  }];
+
+  let assemblyFormat = "attr-dict `:` type($result)";
+  let verifier = [{ return success(); }];
+}
+
+def GPU_NumSubgroupsOp : GPU_Op<"num_subgroups", [NoSideEffect]>,
+    Arguments<(ins)>, Results<(outs Index:$result)> {
+  let description = [{
+    Returns the number of subgroups within a workgroup.
+
+    Example:
+
+    ```mlir
+    %numSg = gpu.num_subgroups : index
+    ```
+  }];
+
+  let assemblyFormat = "attr-dict `:` type($result)";
+  let verifier = [{ return success(); }];
+}
+
+def GPU_SubgroupSizeOp : GPU_Op<"subgroup_size", [NoSideEffect]>,
+    Arguments<(ins)>, Results<(outs Index:$result)> {
+  let description = [{
+    Returns the number of threads within a subgroup.
+
+    Example:
+
+    ```mlir
+    %sgSz = gpu.subgroup_size : index
+    ```
+  }];
+
+  let assemblyFormat = "attr-dict `:` type($result)";
+  let verifier = [{ return success(); }];
+}
+
 def GPU_GPUFuncOp : GPU_Op<"func", [HasParent<"GPUModuleOp">,
                                     AutomaticAllocationScope, FunctionLike,
                                     IsolatedFromAbove, Symbol]> {
diff --git a/mlir/lib/Conversion/GPUToSPIRV/ConvertGPUToSPIRV.cpp b/mlir/lib/Conversion/GPUToSPIRV/ConvertGPUToSPIRV.cpp
--- a/mlir/lib/Conversion/GPUToSPIRV/ConvertGPUToSPIRV.cpp
+++ b/mlir/lib/Conversion/GPUToSPIRV/ConvertGPUToSPIRV.cpp
@@ -67,6 +67,18 @@
                   ConversionPatternRewriter &rewriter) const override;
 };
 
+/// Pattern lowering subgoup size/id to loading SPIR-V invocation
+/// builtin variables.
+template <typename SourceOp, spirv::BuiltIn builtin>
+class SingleDimLaunchConfigConversion : public SPIRVOpLowering<SourceOp> {
+public:
+  using SPIRVOpLowering<SourceOp>::SPIRVOpLowering;
+
+  LogicalResult
+  matchAndRewrite(SourceOp op, ArrayRef<Value> operands,
+                  ConversionPatternRewriter &rewriter) const override;
+};
+
 /// This is separate because in Vulkan workgroup size is exposed to shaders via
 /// a constant with WorkgroupSize decoration. So here we cannot generate a
 /// builtin variable; instead the information in the `spv.entry_point_abi`
@@ -276,6 +288,16 @@
   return success();
 }
 
+template <typename SourceOp, spirv::BuiltIn builtin>
+LogicalResult
+SingleDimLaunchConfigConversion<SourceOp, builtin>::matchAndRewrite(
+    SourceOp op, ArrayRef<Value> operands,
+    ConversionPatternRewriter &rewriter) const {
+  auto spirvBuiltin = spirv::getBuiltinVariableValue(op, builtin, rewriter);
+  rewriter.replaceOp(op, spirvBuiltin);
+  return success();
+}
+
 LogicalResult WorkGroupSizeConversion::matchAndRewrite(
     gpu::BlockDimOp op, ArrayRef<Value> operands,
     ConversionPatternRewriter &rewriter) const {
@@ -457,5 +479,11 @@
       LaunchConfigConversion<gpu::GridDimOp, spirv::BuiltIn::NumWorkgroups>,
       LaunchConfigConversion<gpu::ThreadIdOp,
                              spirv::BuiltIn::LocalInvocationId>,
+      SingleDimLaunchConfigConversion<gpu::SubgroupIdOp,
+                                      spirv::BuiltIn::SubgroupId>,
+      SingleDimLaunchConfigConversion<gpu::NumSubgroupsOp,
+                                      spirv::BuiltIn::NumSubgroups>,
+      SingleDimLaunchConfigConversion<gpu::SubgroupSizeOp,
+                                      spirv::BuiltIn::SubgroupSize>,
       TerminatorOpConversion, WorkGroupSizeConversion>(context, typeConverter);
 }
diff --git a/mlir/lib/Dialect/SPIRV/SPIRVLowering.cpp b/mlir/lib/Dialect/SPIRV/SPIRVLowering.cpp
--- a/mlir/lib/Dialect/SPIRV/SPIRVLowering.cpp
+++ b/mlir/lib/Dialect/SPIRV/SPIRVLowering.cpp
@@ -551,6 +551,16 @@
         builder.create<spirv::GlobalVariableOp>(loc, ptrType, name, builtin);
     break;
   }
+  case spirv::BuiltIn::SubgroupId:
+  case spirv::BuiltIn::NumSubgroups:
+  case spirv::BuiltIn::SubgroupSize: {
+    auto ptrType = spirv::PointerType::get(builder.getIntegerType(32),
+                                           spirv::StorageClass::Input);
+    std::string name = getBuiltinVarName(builtin);
+    newVarOp =
+        builder.create<spirv::GlobalVariableOp>(loc, ptrType, name, builtin);
+    break;
+  }
   default:
     emitError(loc, "unimplemented builtin variable generation for ")
         << stringifyBuiltIn(builtin);
diff --git a/mlir/test/Conversion/GPUToSPIRV/builtins.mlir b/mlir/test/Conversion/GPUToSPIRV/builtins.mlir
--- a/mlir/test/Conversion/GPUToSPIRV/builtins.mlir
+++ b/mlir/test/Conversion/GPUToSPIRV/builtins.mlir
@@ -178,3 +178,51 @@
     }
   }
 }
+
+// -----
+
+module attributes {gpu.container_module} {
+  // CHECK-LABEL:  spv.module Logical GLSL450
+  // CHECK: spv.globalVariable [[SUBGROUPID:@.*]] built_in("SubgroupId")
+  gpu.module @kernels {
+    gpu.func @builtin_subgroup_id() kernel
+      attributes {spv.entry_point_abi = {local_size = dense<[16, 1, 1]>: vector<3xi32>}} {
+      // CHECK: [[ADDRESS:%.*]] = spv._address_of [[SUBGROUPID]]
+      // CHECK-NEXT: {{%.*}} = spv.Load "Input" [[ADDRESS]]
+      %0 = gpu.subgroup_id : index
+      gpu.return
+    }
+  }
+}
+
+// -----
+
+module attributes {gpu.container_module} {
+  // CHECK-LABEL:  spv.module Logical GLSL450
+  // CHECK: spv.globalVariable [[NUMSUBGROUPS:@.*]] built_in("NumSubgroups")
+  gpu.module @kernels {
+    gpu.func @builtin_num_subgroups() kernel
+      attributes {spv.entry_point_abi = {local_size = dense<[16, 1, 1]>: vector<3xi32>}} {
+      // CHECK: [[ADDRESS:%.*]] = spv._address_of [[NUMSUBGROUPS]]
+      // CHECK-NEXT: {{%.*}} = spv.Load "Input" [[ADDRESS]]
+      %0 = gpu.num_subgroups : index
+      gpu.return
+    }
+  }
+}
+
+// -----
+
+module attributes {gpu.container_module} {
+  // CHECK-LABEL:  spv.module Logical GLSL450
+  // CHECK: spv.globalVariable [[SUBGROUPSIZE:@.*]] built_in("SubgroupSize")
+  gpu.module @kernels {
+    gpu.func @builtin_subgroup_size() kernel
+      attributes {spv.entry_point_abi = {local_size = dense<[16, 1, 1]>: vector<3xi32>}} {
+      // CHECK: [[ADDRESS:%.*]] = spv._address_of [[SUBGROUPSIZE]]
+      // CHECK-NEXT: {{%.*}} = spv.Load "Input" [[ADDRESS]]
+      %0 = gpu.subgroup_size : index
+      gpu.return
+    }
+  }
+}
diff --git a/mlir/test/Dialect/GPU/ops.mlir b/mlir/test/Dialect/GPU/ops.mlir
--- a/mlir/test/Dialect/GPU/ops.mlir
+++ b/mlir/test/Dialect/GPU/ops.mlir
@@ -44,6 +44,10 @@
       %gDimY = "gpu.grid_dim"() {dimension = "y"} : () -> (index)
       %gDimZ = "gpu.grid_dim"() {dimension = "z"} : () -> (index)
 
+      %sgId = gpu.subgroup_id : index
+      %numSg = gpu.num_subgroups : index
+      %SgSi = gpu.subgroup_size : index
+
       %one = constant 1.0 : f32
       %sum = "gpu.all_reduce"(%one) ({}) {op = "add"} : (f32) -> (f32)